COMPSCI 223: Computational Microeconomics

1. COMPSCI 223:Computational Microeconomics Instructor: Vincent Conitzer (Kimberly J. Jenkins University Professor of New TechnologiesProfessor of Computer Science, Professor of Economics, and Professor of Philosophy) conitzer@cs.duke.edu https://www2.cs.duke.edu/courses/spring18/compsci223/ TAs: Harsh Parikh and Hanrui Zhang

2. http://econ.cs.duke.edu CS-ECON

4. computer architecture (von Neumann architecture) Computer Science & Engineering History game theory (minimax theorem) ? Economic Theory linear programming (duality) Mathematical Optimization & Operations Research John von Neumann 1950 2000 1900

5. What is Economics? • “the social science that studies the production, distribution, and consumption of goods and services.” [Wikipedia, Jan. 2018] • Some key concepts: • Economic agents or players (individuals, households, firms, bots, …) • Agents’ current endowments of goods, money, skills, … • Possible outcomes ((re)allocations of resources, tasks, …) • Agents’ preferences or utility functions over outcomes • Agents’ beliefs (over other agents’ utility functions, endowments, production possibilities, …) • Agents’ possible decisions/actions • Mechanism that maps decisions/actions to outcomes

6. An economic picture v( ) = 200 $ 800 v( ) = 200 v( ) = 100 v( ) = 400 v( ) = 400 $ 600 $ 200

7. After trade (a more efficient outcome) … but how do we get here? Unstructured trade? Auctions? Exchanges? v( ) = 200 $ 1100 v( ) = 200 v( ) = 100 v( ) = 400 v( ) = 400 $ 400 $ 100

8. Some distinctions in economics • Descriptive vs. normative economics • Descriptive: • seeks only to describe real-world economic phenomena • does not care if this is in any sense the “right” outcome • Normative: • studies how people “should” behave, what the “right” or “best” outcome is • Microeconomics vs. macroeconomics • Microeconomics: analyzes decisions at the level of individual agents • deciding which goods to produce/consume, setting prices, … • “bottom-up” approach • Macroeconomics: analyzes “the sum” of economic activity • interest rates, inflation, growth, unemployment, government spending, taxation, … • “big picture”

9. What is Computer Science? • “the study of automating algorithmic processes that scale. A computer scientist specializes in the theory of computation and the design of computational systems.” [Wikipedia, Jan. 2018] • A computational problem is given by a function f mapping inputs to outputs • For integer x, let f(x) = 0 if x is prime, 1 otherwise • For initial allocation of resources + agent utilities x, let f(x) be the (re)allocation that maximizes the sum of utilities • An algorithm is a fully specified procedure for computing f • E.g., sieve of Eratosthenes • A correct algorithm always returns the right answer • An efficient algorithm returns the answer fast • Computer science is also concerned with building larger artifacts out of these building blocks (e.g., personal computers, spreadsheets, the Internet, the Web, search engines, artificial intelligence, …)

10. Resource allocation as a computational problem (Part 1 of the course) input output v( ) = $400 v( ) = $600 $ 750 $ 800 v( ) = $500 $ 450 v( ) = $400 Here, gains from trade ($300) are divided evenly (not essential) $ 400

11. Economic mechanisms agents’ bids “true” input result ) = $500 v( ) = $400 v( v( ) = $501 v( ) = $600 agent 1’s bidding algorithm $ 800 exchange mechanism (algorithm) $ 800 $ 800 $ 400 v( ) = $500 ) = $451 v( v( ) = $400 v( ) = $450 agent 2’s bidding algorithm Exchange mechanism designer does not have direct access to agents’ private information Agents will selfishly respond to incentives $ 400 $ 400

12. Game theory (Part 2 of the course) • Game theory studies settings where agents each have • different preferences (utility functions), • different actions that they can take • Each agent’s utility (potentially) depends on all agents’ actions • What is optimal for one agent depends on what other agents do • Very circular! • Game theory studies how agents can rationally form beliefs over what other agents will do, and (hence) how agents should act • Useful for acting as well as predicting behavior of others

13. Penalty kick example probability .7 probability .3 action probability 1 Is this a “rational” outcome? If not, what is? action probability .6 probability .4

14. Mechanism design (Part 3 of the course) • Mechanism = rules of auction, exchange, … • A function that takes reported preferences (bids) as input, and produces outcome (allocation, payments to be made) as output f( )= ) = $500 v( ) = $400 v( v( ) = $400 v( ) = $600 $ 750 $ 400 $ 800 $ 450 • The entire function f is one mechanism • E.g., the mechanism from part 1: find allocation that maximizes (reported) utilities, distribute (reported) gains evenly • Other mechanisms choose different allocations, payments

15. Example: (single-item) auctions • Sealed-bid auction: every bidder submits bid in a sealed envelope • First-price sealed-bid auction: highest bid wins, pays amount of own bid • Second-price sealed-bid auction: highest bid wins, pays amount of second-highest bid bid 1: $10 first-price: bid 1 wins, pays $10 second-price: bid 1 wins, pays $5 bid 2: $5 bid 3: $1 0

16. Which auction generates more revenue? • Each bid depends on • bidder’s true valuation for the item (utility = valuation - payment), • bidder’s beliefs over what others will bid (→ game theory), • and... the auction mechanism used • In a first-price auction, it does not make sense to bid your true valuation • Even if you win, your utility will be 0… • In a second-price auction, (we will see later that) it always makes sense to bid your true valuation bid 1: $10 a likely outcome for the first-price mechanism a likely outcome for the second-price mechanism bid 1: $5 bid 2: $5 bid 2: $4 bid 3: $1 bid 3: $1 0 0 Are there other auctions that perform better? How do we know when we have found the best one?

17. Mechanism design… • Mechanism = game • → we can use game theory to predict what will happen under a mechanism • if agents act strategically • When is a mechanism “good”? • Should it result in outcomes that are good for the reported preferences, or for the true preferences? • Should agents ever end up lying about their preferences (in the game-theoretic solution)? • Should it always generate the best allocation? • Should agents ever burn money?(!?) • Can we solve for the optimal mechanism?

18. How are we going to solve these problems? (Part 0) • This is not a programming course • Will use optimization software • GNU Linear Programming Kit (GLPK) • Linear programming, mixed integer linear programming

19. Uses of LP, MIP in this course

20. Other settings/applications

21. Combinatorial auctions (in Part 1) Simultaneously for sale: , , bid 1 v( ) = $500 bid 2 v( ) = $700 bid 3 v( ) = $300 used in truckload transportation, industrial procurement, radio spectrum allocation, …

22. Voting (in Part 1) voting rule (mechanism) determines winner based on votes > > • Can vote over other things too • Where to go for dinner tonight, other joint plans, … • Many different rules exist for selecting the winner > >

23. Kidney exchange (in Part 1) donor 1 (patient 1’s friend) patient 1 compatibilities donor 2 (patient 2’s friend) patient 2 donor 3 (patient 3’s friend) patient 3 donor 4 (patient 4’s friend) patient 4

24. Game playing & AI (in Part 2) perfect information games: no uncertainty about the state of the game (e.g. tic-tac-toe, chess, Go) imperfect information games: uncertainty about the state of the game (e.g., poker) “nature” white 1 gets King 1 gets Jack Qa1-a8 Qa1-f6 player 1 player 1 black black raise raise check check Kf8-f7 Kf8-g7 Kf8-g8 Kf8-e8 player 2 player 2 black wins white wins draw draw call fold call fold call fold call fold • Optimal play: value of each node = value of optimal child for current player (backward induction, minimax) • For chess and Go, tree is too large • Use other techniques (heuristics, limited-depth search, alpha-beta, deep learning, …) • Top computer programs better than humans in chess, not yet in Go 2 1 1 1 -2 1 -1 1 • Player 2 cannot distinguish nodes connected by dotted lines • Backward induction fails; need more sophisticated game-theoretic techniques for optimal play • Small poker variants can be solved optimally • Humans still better than top computer programs at full-scale poker (at least most versions) • Top computer (heads-up) poker players are based on techniques for game theory ?

25. Real-world security applications (in Part 2) Milind Tambe’s TEAMCORE group (USC) • Airport security • Where should checkpoints, canine units, etc. be deployed? • Federal Air Marshals • Which flights get a FAM? • US Coast Guard • Which patrol routes should be followed? • Wildlife Protection • Where to patrol to catch poachers or find their snares?

26. Prediction markets

27. Financial securities (in Part 1) • Tomorrow there must be one of • Agent 1 offers $5 for a security that pays off $10 if or • Agent 2 offers $8 for a security that pays off $10 if or • Agent 3 offers $6 for a security that pays off $10 if • Can we accept some of these at offers at no risk?

28. How to incentivize a weather forecaster (in Part 3) • Forecaster’s bonus can depend on • Prediction • Actual weather on predicted day • Reporting true beliefs should maximize expected bonus P( ) = .5 P( ) = .8 P( ) = .3 P( ) = .1 P( ) = .2 P( ) = .1

29. Sponsored search / ad auctions (in Part 3) • Choice of ads (if any) to show determined by: • Advertiser bid • Predicted likelihood of click